Jake Thomas

I am currently a PhD student in the Mathematics of Real World Systems II Centre for Doctoral Training. My PhD is focused on the application of possibility theory to machine learning, with a particular focus on reinforcement learning.

Beyond my PhD, I am a machine learning researcher at Advai, where I lead the agentic AI research and development.

Research Interests

• Machine Learning
• Reinforcement Learning
• Possibility Theory

Papers

• Downer, G., Craven, S., Ruck, D., & Thomas, J. (2025). Text2VLM: Adapting Text-Only Datasets to Evaluate Alignment Training in Visual Language Models. arXiv preprint arXiv:2507.20704.
• Thomas, J., & Houssineau, J. (2024). Improving Active Learning with a Bayesian Representation of Epistemic Uncertainty. arXiv preprint arXiv:2412.08225.

Education

• Mathematics of Systems (MSc), 2019-2020, University of Warwick, Distinction
  Research Project: Multi-armed Bandit Problems and Possibility Theory
  Supervisor: Dr Jeremie Houssineau

• Master of Mathematics (MMath), 2014-2019, University of Warwick, 1st Class
  MMath Project: 'The Stochastic Exit Problem for Irreversible Systems’
  Supervisor: Dr Hendrik Weber (University of Bath)